| CARVIEW |
JJ (Jeong Joon) Park
jjparkcv (at) umich (dot) edu
I'm an assisant professor at the University of Michigan CSE (Office: Leinweber 3154).
I'm broadly interested in computer vision, graphics, and artificial intelligence. My current research focus is on 3D/4D reconstruction and generative modeling and their applications to robotics, medical imaging, and scientific problems.
I'm looking for students and postdoc applicants!
Please refer to the note below for details. I encourage interested students to apply to the UMichigan CSE PhD program and mention my name in the application.
Prospective Students and Postdocs
I'm looking for PhD and postdoc applicants who have research experience with (but not limited to):- Geometric AI
- Graphics or scientific simulation
- Computer vision (including medical imaging) and generative models
- Robot learning
- Machine learning
- Computational Neuroscience
For UMichigan undergrads and masters' students please send an email with resume -- note that I expect a significant time commitment (>15hrs/week). Unfortunately, I will not be able to respond to all emails.
Current Students
Ang Cao (2020-, co-advised with Justin Johnson)Liam Wang (2024-, NSF GRFP Fellow)
Zichen Wang (2024- )
Lixuan Chen (2024-, Co-advised with Liyue Shen)
Hyelin Nam (2025-)
Yunhao Luo (2025-, Co-advised with Nima Fazeli)
Ling Xiao (2025-, Co-advised with Karthik Duraisamy)
Publications
CurveCloudNet: Processing Point Clouds with
1D Structure
C. Stearns, D. Rempe, A. Fu, J. Liu, S. Mascha, JJ Park, D. Paschalidou, L. Guibas
CVPR 2024
Summary
We introduce a new point cloud processing scheme which takes advantage of the curve-like structure inherent to modern depth sensors. While existing backbones discard the rich 1D traversal patterns, CurveCloudNet parameterizes the point cloud as a collection of polylines (a "curve cloud”), establishing a local surface-aware ordering on the points.
GeNVS: Generative Novel View Synthesis with
3D-Aware Diffusion Models
E. Chan*, K Nagano*, M Chan*, A. Bergman*, JJ Park*,
A. Levy, M. Aittala, S. Mello, T. Karras, G. Wetzstein
ICCV 2023 (Oral)
Summary
We present a diffusion model for 3D-aware generative novel view synthesis from as few as a single input image. Our model samples from the distribution of possible renderings consistent with the input and is capable of rendering plausible novel views of unbounded regions.
CC3D: Layout-Conditioned Generation of Compositional 3D Scenes
Sherwin Bahmani*,
Jeong Joon Park*,
Despoina Paschalidou,
Xingguang Yan,
Gordon Wetzstein,
Leonidas Guibas,
Andrea Tagliasacchi
ICCV 2023
Summary
We introduce a conditional generative model that synthesizes complex 3D scenes conditioned on 2D semantic scene layouts. Different from most
existing 3D GANs that operate on aligned
single objects, we focus on generating complex 3D scenes
with multiple objects, by modeling the compositional nature
of 3D scenes.
Seeing the World in a Bag of Chips
Jeong Joon Park, Aleksander Holynski, Steve Seitz
CVPR 2020 (Oral)
WIRED Scientific American
